This is the project page for Maximum Classifier Discrepancy. The work was accepted by CVPR 2018 Oral. [Paper Link(arxiv)].

Abstract

We propose a new approach for unsupervised domain adaptation, which attempts to align distributions of source and target by utilizing the task-specific decision boundaries. We propose to maximize the discrepancy between two classifiers’ outputs to detect target samples that are far from the support of the source. A feature generator learns to generate target features near the support to minimize the discrepancy. Our method is applicable to classification and semantic segmentation. The implementation is availble now !

Example

Our method demonstrates good performance both on classification and semantic segmentation for unsupervised domain adaptation. The examples of semantic segmentation are shown here.



Codes

[Classification] [Segmentation]

Citation

If you use this code for your research, please cite our papers (This will be updated when cvpr paper is publicized).

@article{saito2017maximum,
  title={Maximum Classifier Discrepancy for Unsupervised Domain Adaptation},
  author={Saito, Kuniaki and Watanabe, Kohei and Ushiku, Yoshitaka and Harada, Tatsuya},
  journal={arXiv preprint arXiv:1712.02560},
  year={2017}
}